Part one ended on a wrong note in a tidy row, and a wrong note sounds cheap. You look it up, you find it is out of date, you fix it, you carry on. Ten minutes. Everyone has had worse mornings.
That is the accounting most writers do, and it is wrong in an interesting way: it prices the moment you catch the error. Almost none of the cost lives there. The cost lives in the interval between the entry going stale and you finding out, and in what you did during that interval on the strength of a fact that was not true.
The charge is deferred, and interest accrues
Consider what happens when the plan says the sister is thirty-four and the draft, since chapter nine, has quietly made her the elder.
You do not find out in chapter ten. You find out in chapter twenty-six, or in revision, or from a reader — because the whole reason you wrote the entry down was to stop checking, and it worked. The entry did its job. It answered every time you asked, instantly, without making you reread anything. It just answered wrongly.
So between chapter nine and the moment of discovery, every scene that touched her age was written on top of a false premise. Not necessarily broken — often just slightly wrong in the way a hem is wrong, pulling on things a few pages away. The inheritance argument. Who was old enough to remember the fire. The line where he calls her kid and it is meant to sting.
That is the shape of the real cost. It is not the ten minutes of correction. It is a repair whose size is set by how long the error was allowed to be load-bearing, and you do not choose that number. The plan chooses it, by being silently confident until something finally contradicts it loudly enough to notice.
You pay for a stale entry at the moment you discover it, but the size of the bill was set months earlier — by how much of the draft got built on top of it while it still looked correct.
Three costs, and the third is the one nobody expects
The first is rework, and it is the smallest. Scenes to revisit, lines to adjust, a timeline to redraw. Real work, sometimes a fortnight of it, but bounded and visible. You can see it, plan around it, and know when it is done. If this were the only cost I would not have given it a part of its own.
The second is the decisions you make twice. This one hides better. When a writer stops trusting their notes — and the five-entry test in part one tends to produce exactly that — they do not usually announce it. They just start re-deciding small things at the keyboard rather than looking them up, because looking up now costs a verification step and re-deciding is free.
The trouble is that re-deciding is only free the first time. Decide the shop's opening hours in chapter twelve because you cannot be bothered to confirm them, and you have not resolved the question; you have created a second answer. Now there are two, both in the manuscript, neither marked, and the version in the notes is a third. You have quietly converted a lookup problem into a continuity problem, and continuity problems are found by readers.
The third is the one that actually stops books, and it is a loss of nerve.
Once you know your own reference is unreliable, you stop being able to write forward at speed — because every fact you lean on now carries a small unpaid debt of verification.
This is worth sitting with, because it is the mechanism that turns a filing annoyance into a stalled manuscript. Drafting at any pace depends on being able to assert things. She was nineteen that summer. The road runs east. He has never been in the house before. If those assertions are cheap, the scene moves. If each one costs a decision about whether to trust yourself, the scene does not move — and the writer experiences that as being blocked, or as the book being difficult, or as their own flagging discipline.
It is none of those. It is a reference system that has lost its authority, being felt as a loss of momentum. And it is one of the few writing problems that gets strictly worse the further you are into the book, because there is more draft above you and more of it you cannot remember writing.
Why this lands hardest on the drafts closest to finishing
There is a cruelty in the timing that is worth naming, because it inverts what most advice assumes.
At twenty thousand words your plan is roughly current. You wrote most of it recently, you remember the decisions, and you can hold the book in your head on a walk. The reference is barely load-bearing, so its errors are cheap.
At ninety thousand it is the opposite on every count. The plan is months old in places. The draft is far too large to hold. You are relying on the notes constantly and checking them rarely, and the accumulated stack of overtaken facts — the ones nobody ever decided, from part one — is at its deepest. The reference matters most exactly when it is least trustworthy.
Which means this is not a beginner's problem that you grow out of. It is a late-draft problem that grows into you, and it arrives at the point where abandoning the book is most tempting and most costly.
Worse than no notes at all?
In one specific respect, yes, and it is worth being exact rather than provocative about it.
A writer with no notes knows they have no notes. Their uncertainty is calibrated: they are aware that any fact they cannot recall must be looked up in the draft, and they either do it or knowingly take the risk. The cost is high but it is visible, and visible costs get managed.
A writer with an out-of-date reference has uncertainty that is miscalibrated, which is a different and worse condition. They feel confident on precisely the facts most likely to be wrong, because confidence is what the entry supplies. The whole error class is invisible until something breaks.
Which is why the fix is not diligence. You cannot correct a belief you have no reason to doubt.
Take the entries the five-entry test marked superseded or overtaken, and for each one ask a single question: what is the earliest scene that would have to change if this entry were corrected? The distance between that scene and where you are now is your actual exposure. It is usually the first time this stops feeling like tidying.
What would have to be true instead
Everything above comes from one structural fact, and it is worth stating plainly before part three, because it tells you what a fix has to do.
The plan and the draft are two separate documents. Only one of them is updated by the act of writing. The other is updated by an act of maintenance — a deliberate, unpaid, easy-to-skip administrative task that competes directly with the writing session and reliably loses.
So no amount of discipline solves this, and no better template solves it either. Asking a writer to hand-maintain a second document in perfect sync with the first, for a year, while drafting, is asking for something nobody sustains. The interesting question is not how to be more diligent. It is what the plan would have to be, structurally, for the drift to stop being possible.
That is part three.
